RizzitGO Open API Added Keyword Search: What Changed in September 2026

The Open API changelog says version 2.10 introduced POST /open-api/goods/search under the goods:search:read scope. A request contains a keyword, page number and source code, with an optional sort value. The documented sort states are relevance, price descending, price ascending and sales descending.

This is a partner API protected by OAuth2 client credentials, tenant isolation and granted scopes. It should not be described as an anonymous public feed or proof that every RizzitGO user interface exposes the same controls. The reliable claim is simply that approved partner applications can request structured keyword results through the documented endpoint.

The response model includes a goods ID, title, detail URL, main image, source, price, original price, sales count and discount fields. Pagination adds the current page, page size, total pages and total result count. Those fields give a catalog index enough structure to group and compare results without reducing every row to a pasted marketplace URL.

The source field is especially important. The keyword endpoint documents Taobao, 1688, Weidian and Xianyu as distinct sources. The same-looking item can therefore appear under different marketplaces, sellers or listing records. A responsible index should preserve that context rather than presenting visually similar results as one verified product.

Traditional shared spreadsheets are easy to distribute, but their rows age quickly. A structured endpoint makes it possible for an approved integration to refresh titles, images and prices while keeping the presentation searchable. It also makes snapshot dates more important: once a result is copied into an independent directory, the copy can drift from its source.

This is why product counts should be treated as dated measurements rather than permanent claims. An index can remove duplicates, exclude incomplete records or refresh a changed listing and legitimately show a different total from another directory using a different pull time and normalization method.

The endpoint does not return an unlimited catalog in one response. Its model reports the current page, the returned size, total pages and total count. A partner application that wants broader coverage has to continue requesting pages and decide what to do when a later page changes during the same collection run.

That matters because marketplace search is not a frozen database export. New listings can appear, prices can change and ranking can move while pages are being collected. If an index refreshes several pages at different times, the final directory may contain a mixture of moments unless the publisher records a clear pull window and repeats failed pages consistently.

Pagination also explains why a headline count and the number of visible cards may differ. The API total describes the source response for that query and source; the published index may display only records that pass image, title, category and duplicate checks.

The four documented states answer different questions. Relevance attempts to order results around the entered words, price sorting rearranges the same search around a numeric field, and sales sorting emphasizes the source marketplace's sales value. None of those choices is described as a quality, authenticity or reliability ranking.

This distinction becomes easy to lose on a spreadsheet site. A row near the top can look endorsed simply because it is prominent. A transparent directory should identify whether order comes from the source, from a selected sort mode or from its own normalization rules instead of calling the first result the best find.

The version history is useful for the same reason. It documents when keyword search entered the partner API, but it does not say that every older spreadsheet was rebuilt or that every consumer-facing page changed on that date.

The search response does not authenticate products, guarantee inventory or certify delivered quality. The documentation also says member-specific collection status and personalized share links are not returned. Public fields are useful discovery signals, but they do not replace the current product page, seller terms or warehouse evidence.

This independent finder currently uses its own saved catalog snapshot; it is not connected to the RizzitGO Open API. The RizzitGO documentation is analyzed here because it is a primary source for understanding the platform's published data model—not because the two datasets are interchangeable.